Pomegranate is the fruit of the pomegranate tree (Punica granatum), which is consumed worldwide and is currently cultivated in Iran, Spain, Italy, Afghanistan, the United States, India, China, Russia, Uzbekistan, Morocco, Greece and Mexico (^{Koba and Yanagita, 2011}). Its production has gained importance due to the functional properties it possesses, which is why various food products such as juices and liquors are produced, in addition to its importance in the cosmetic industry (^{Ge et al., 2021}). In Mexico in 2021, 1,251 ha were cultivated and 8,636 t were produced. The states with the highest harvest volumes were Morelos, Hidalgo and Oaxaca with 1,622, 1,467 and 1,327 t, respectively (^{SIAP, 2021}), destined for both domestic consumption and export. Pomegranate orchards are affected by diseases induced by various microorganisms such as Alternaria spp. and Aspergillus spp., with the greatest impact being those that directly affect the fruit pre-harvest (^{Behzad et al., 2020}). One such disease is gray mold induced by Botrytis cinerea, which has been reported mainly in pomegranate orchards in Greece and Pakistan (^{Bardas et al., 2009}; ^{Alam et al., 2018}). It occurs after flowering at the beginning of fruit formation and is characterized by the presence of spots that increase in size, with light to dark brown expanded lesions of soft consistency, followed by the appearance of gray mycelium on infected surfaces. Fruit may remain mummified on the tree. Recently, this phytosanitary problem has been reported in the State of Mexico, Mexico (^{Patricio-Hernández et al., 2023}). Diagrammatic scales could be a useful tool to estimate the severity of gray mold in pomegranate fruits pre- and post-harvest. These tools allow correct interpretation of disease advancement and progress in crops, defined as sets of illustrations of plants or plant organs, with signs and symptoms that show the percentages of area affected by the disease. This is based on the Weber-Fechner principle, which establishes classes in a logarithmic system that eliminates arbitrary designation of severity levels (^{French and Hebert, 1980}). The scales should be quick and simple when used under field and post-harvest conditions, as well as accurate, precise and reproducible (^{Richard et al., 2021}; ^{Vereschuk et al., 2022}). To date, no scales for assessing the severity of gray mold on pomegranate fruits have been reported. The objective of the present study was to develop and validate a diagrammatic scale to help growers and technicians assess disease severity.

During the months of July to September 2022, 120 pomegranate fruits with and without symptoms of gray mold (Botrytis cinerea) were collected in active production plots located in the municipalities of Chilcuahutla and Taxquillo, Hidalgo, Mexico (20° 18’ 11’’ N, 99° 14’ 23’’ W, 20° 32’ 01’’ N, 99° 20’ 03’’ W, respectively). Subsequently, 60 fruits representative of the different degrees of damage were selected. To obtain the actual severity, each fruit was divided in half in order to photograph the total surface of each one using a Canon T7 camera (Verechuk et al., 2022).

To eliminate the background, the images were processed with GIMP^® v.2.10.12 software. Quantification of the total and affected area was performed using Image Tool v1.8.0. With the obtained data, the actual severity percentage was calculated using the following formula: severity = (diseased area/total image area) * 100 (^{Nutter Jr et al., 2006}; ^{Ortega-Acosta et al., 2016}). The data were used to define the minimum and maximum values of actual severity, which were then used to generate a logarithmic scale with six classes, using the 2LOG v.1 software (^{Mora-Aguilera and Acevedo-Sánchez, 2018}). This follows the Weber-Fechner visual acuity law (^{Horsfall and Cowling, 1978}). The obtained data were used to construct the diagrammatic scale with Adobe Photoshop software (^{Fantin et al., 2018}).

To validate the diagrammatic scale, 60 digital images representative of different degrees of severity were randomly inserted into individual slides to be visualized in Microsoft 365^® PowerPoint and presented to 18 evaluators with and without experience in observing plant diseases. They carried out independent evaluations with approximately 20 s per image for visualization. Data from this first evaluation were expressed as percentage of severity (^{Fragoso-Benhumea et al., 2022}). For the first and second evaluations using the scale, 12 evaluators were selected based on the number of correct scores in the evaluation without the scale and their willingness for subsequent participation (^{Belan et al., 2014}). Each evaluation was carried out with an interval of 7 days between them.

To quantify the accuracy of the severity evaluations made by the evaluators, a simple linear regression was performed to verify the following hypotheses: for the intercept (β ₀) the null hypothesis H₀: β ₀=0 versus H1: β ₀≠0 and for the slope coefficient (β ₁) H₀: β ₁=1 versus H₁: β ₁≠1, with a significance level of 5%, using a t-test. The actual values obtained were used as the independent variable and the estimated values per evaluator were used as the dependent variable (^{Da silva et al., 2019}). This took into account that if the estimated values of the slope differ from 0, they indicate overestimation of the real severity when β ₀ >0 and underestimation if β ₀<0. Similarly, if the slope data differ from 1, they indicate overestimation of disease (>1) or underestimation (<1) at all disease severity levels (^{Nutter Jr. and Schultz, et al., 1995}; ^{Nutter et al., 2006}; ^{Ortega-Acosta et al., 2016}).

Additionally, the precision of the estimation was determined by the coefficient of determination (r²) of the linear regression and the absolute error was plotted. Furthermore, a paired data analysis per evaluator was performed. Statistical analyses were conducted using the Rstudio program (http://www.rstudio.com/).

From field collections in the municipalities of Chilcuautla and Taxquillo, Hidalgo, 120 fruits were obtained, of which 60 were selected based on their actual degree of severity. Those with 0% severity were considered healthy, while those showing signs and symptoms of gray mold were characterized by the presence of brown lesions originating at the base of the calyx and advancing towards the peduncle, causing rotting of the infected area with values ranging from 5 to 100%.

Based on the percentages of damaged area of the 60 selected fruits, the 2LOG program allowed the definition of six severity classes. The ranges and midpoints of each class (0, 1, 2, 3, 4 and 5) were expressed as percentage of affected area: Class 0= 0%, Class 1 = (>1 - 5 - 10)%, Class 2 = (>11 - 25 - 50)%, Class 3 = (>51 - 75 - 85)%, Class 4 = (>86 - 90 - 95)% and Class 5 = (>96 - 100)% (Figure 1).

Figure 1 TableDiagrammatic scale for evaluation of gray mold (Botrytis cinerea) severity on pomegranate (Punica granatum) fruit. Severity intervals for each class are shown in parentheses. 

The accuracy of the evaluations showed significant differences with and without the use of the designed scale. When the scale was not used, the r² values ranged from 0.06 to 0.87 with a mean of 0.44. However, with the use of the scale, the results were from 0.71 to 0.93 for the first evaluation and from 0.70 to 0.97 for the second evaluation, with mean values of 0.81 and 0.90, respectively. Therefore, the estimations were accurate (Table 1).

Regarding the linear regression results, in the evaluation where the scale was not used, the intercept values were greater than one, indicating overestimation of severity by most evaluators. In the case of the slope values, for evaluator 10 (EV10) it was significantly different from 1, however, there was a tendency toward underestimation, with the exception of EV1 and EV11 that showed values closer to 1 (Table 1).

In the initial evaluation employing the scale, the intercept values showed a tendency to overstate disease severity. Specifically, EV3’s intercept significantly deviated from 0. As for the slope values, EV4, EV5, EV6 and EV10 showed overestimation compared to the other evaluators. In the second evaluation, EV4 and EV5 demonstrated an underestimation of severity, as reflected by values exceeding 1. Meanwhile, EV1 and EV7 yielded statistically distinct values from 0. Concerning the slope, none of the data exhibited significant differences, although their values closely approximated 1 (Table 1).

In general, concentrating on the evaluations conducted with the scale and considering the null hypotheses (H₀: β ₀=0 and H₀: β ₁=1), it is observable that β ₀ maintained proximity to 0 across all evaluators. Likewise, β ₁ uniformly hovered around 1. This confirms that using the diagrammatic scale of gray mold severity allows obtaining values of precision and accuracy close to those of real severity, even when there are slight tendencies of underestimation and overestimation. This statement is reinforced when comparing the absolute error values of the evaluations, where a decrease in absolute error is observed when using the designed scale (Figure 2).

Figure 2 Graphical representation of the absolute error of the estimates of gray mold severity in pomegranate fruits. a) Evaluation carried out without support of the diagrammatic scale, b) and c) Evaluations carried out with support of the diagrammatic scale, d) and e) Representation of b) and c) at a smaller scale of absolute error.  

The combinations of the r² values of the evaluations showed that the scales are reproducible. Where the diagrammatic scale was not used, an interval of 0.13 to 0.77, with a mean of 0.44, was obtained. With the scale, the r² intervals were 0.72 to 0.91, and 0.76 to 0.97, in the first and second evaluations, respectively, with an overall mean value of 0.85 (Table 2).

The use of diagrammatic scales is efficient in evaluating disease severity in plants (^{Fantin et al., 2018}). Recently, the presence of gray mold caused by B. cinerea on pomegranate fruits in Mexico was reported (^{Patricio-Hernández et al., 2023}), and to our knowledge, a diagrammatic scale for evaluating the severity of this disease has not been officially reported. The scale developed in the present investigation allows reliable evaluation of the severity of this disease with greater precision and accuracy. When the scale was not used, the average r² value was 0.44, while in the first and second evaluations with the scale, averages of 0.81 and 0.90 were obtained, respectively. This represents a significant increase, agreeing with what was observed by ^{Belan et al. (2014}), who obtained r² values of 0.89 and 0.87 in the first and second evaluations, respectively, when using their designed scale for coffee tree leaf spot.

The r² values of the present investigation were significantly higher in the evaluations where the scale was used. This is similar to those reported by ^{Fragoso-Benhumea et al. (2022}), who obtained values of 0.90 to 0.97, with a mean of 0.93 when evaluating the severity of rust (Uromyces viciae-fabae) in the broad bean crop, with and without the use of diagrammatic scales. They concluded that this value improves with the use of diagrammatic scales. This indicates that the designed scale can be used in the field, since it has precision values (r²) close to 1 in the evaluations.

Several studies have shown a tendency to overestimate (^{Figueiredo et al., 2022}; ^{Pereira et al., 2021}) and to a lesser extent underestimate (^{Braga et al., 2020}) when diagrammatic scales are not used. In the present work, overestimates were observed in most of the absolute data when the diagrammatic scale was not used, in contrast to the evaluations where it was used. This behavior may be due to visual stimuli, such as similar colorations in the lesions that are not considered for the evaluation of severity, as reported by ^{Perina et al. (2019}). They identified factors that can lead to overestimation of the severity of brown spot on leaves of Citrus reticulata caused by Alternaria sp. due to the presence of different pigments on the leaf surface that do not correspond to the evaluated disease. This can affect both experienced and inexperienced evaluators.

The use of the scale allowed a considerable decrease in the absolute error of the evaluations, compared to when it was not used (without the scale it was from -59 to 39%, with the scale it was from -0.4 to 0.4% and -0.16 to 0.48% in the first and second evaluations, respectively). Similar results were obtained by ^{Muños-Arias et al. (2020}) for gray mold on Rubus glaucus, where this value decreased with use of the diagrammatic scale (from -30 to 30% without the scale to -20 to 20% with the scale). On the other hand, ^{Ortega-Acosta et al. (2016}) and ^{Nutter Jr and Schultz (1995}) mention that absolute error values less than 5% are considered acceptable. In this work, results of -0.35 to 0.5% were obtained on average for the two evaluations, which supports the accuracy of the evaluations when using the scale. Thus, the present research provides a reliable instrument for evaluating the severity induced by gray mold in pomegranate fruits.

When r² values among evaluators were compared, it was evident that the designed scale exhibited precision, evident in its capacity to narrow the variation intervals within comparisons when employing the diagrammatic scale. This particular methodology enabled the accurate, precise, and reproducible estimation of gray mold severity attributed to B. cinerea on pomegranate fruits. As a result, the scale holds the potential to serve as an effective tool for disease management, monitoring, and surveillance purposes.